# Why automations don't stick with your team | Auspex > Some abandoned automations should be switched off, not rescued. Four reasons they die, and a review protocol: keep, redesign or switch off. Source: https://auspex.company/en/blog/why-automations-get-abandoned/ [Blog](/en/blog/) / Field Notes # Why some automations stick and others quietly die By: Andrew Maryasov · Aug 12, 2026 · 9 min read An automation nobody uses usually gets blamed on the human factor: the team never got used to it, the manager didn't push hard enough, it's time to train everyone again. Katie Parrott, a staff writer at Every, took apart her own abandoned AI systems and came to a different conclusion: walking away from a tool can be a rational decision when it was built for someone else's problem and competes with a behavior that already solves yours. Auspex checked her analysis against three behavioral science studies — on how long a habit takes to form, on the pull of immediate rewards and on the limits of planning. For a company of 10–200 people the conclusion is uncomfortable: some abandoned automations are better switched off than rescued, and that starts with a review, not with one more training session. ![Four reasons an automation dies: it competes with a working habit, not with a vacuum; a delayed benefit loses to an immediate one; a written procedure doesn't create motivation; the 21-day rule is a myth](/images/blog/why-automations-get-abandoned-en.webp) An abandoned automation is not a failure of discipline ## Why is an abandoned automation not a failure of discipline? In July 2026, Katie Parrott, a staff writer at Every, described her own case. After watching Every CEO Dan Shipper run his workday through Tend, a system built on OpenAI Codex, on a video call, she handed the transcript to Codex the moment the call ended and had it build her own version — Attention Desk. A few days later it was sitting in her pinned chats with an unread marker, and it was never opened again. Her first reaction was predictable: it's me, I wasn't consistent enough. But the list of reasons a work tool goes unused has almost nothing to do with discipline. The task comes up too rarely. The trigger doesn't fire. The output creates more work than it saves. > A tool can fail because my habits don't support it; my habits can also be rational responses to the work I actually do. > > — Katie Parrott, Every A different diagnosis means a different next step. If the cause is discipline, what follows is training, reminders and monitoring. If the cause is design, what follows is a review in which some automations get switched off. ## Why is the 21-day rule a myth, and why does it set the wrong time to check? The most common mistake in evaluation is checking whether an automation has stuck two weeks after launch. That timeframe is a holdover from the 21-day rule, and the rule doesn't come from research: the number was popularized by Maxwell Maltz's book Psycho-Cybernetics, and it described observations of plastic surgery patients, not how work habits form. A systematic review and meta-analysis led by Ben Singh (University of South Australia), published in the journal Healthcare in December 2024, pooled 20 experimental studies with 2,601 participants. Four of them measured the time it took for a habit to appear, and the results diverged: a median of 59 days in the first, 66 days in the second, a mean of 91 days in the third, and means of 106 and 154 days in the fourth, depending on whether the action was performed in the morning or in the evening. The range within the first study was 4 to 335 days, and only 23% of participants reached the set automaticity threshold, so the median is calculated over the minority that got there. Both extremes are equally useless. Evaluating a new automation after two weeks measures nothing. But “give people time” with no criteria doesn't work either: some participants never reach automaticity, however long you give them. ## What carries over from these studies to business processes, and what doesn't? The 2024 review looked at everyday behaviors: drinking water, flossing, physical activity, diet. The authors name the limitations plainly — there are few studies, and 11 of the 20 carry a high risk of bias. You can't transfer these numbers to a CRM literally, and anyone who promises you “59 days to the habit of using the system” is selling fiction. What carries over is the factors, not the timelines. The review names the ones that affect habit strength: how often the action is repeated, how stable the context is, the type of action, how enjoyable it feels and, as a separate item, personal choice. Habits people chose for themselves formed more strongly than habits assigned from outside. One more observation: the first repetitions bring the biggest gain in automaticity, and then comes a point after which more repetitions add nothing. For a business, two things follow. An automation a sales rep asked for has better odds than the same automation handed down from above. And if a process runs a few times a year, there is no stable context for a habit at all, and no number of reminders will change that. ## What is a new automation really competing with? Attention Desk didn't simply fail to stick — it lost a competition. Parrott checks Slack 40–50 times a day, and every check delivers something tangible right away: relief from not having to write, or a dose of fresh news. The peace of mind Attention Desk promised was no match for that kind of reward. In a company of 10–200 people, Slack's role is played by the department head's Excel file, a group chat in a messaging app and the phrase “it's quicker to call than to enter it in the system.” These habits pay too: in control, in speed, in personal contact. A new automation has to do more than be objectively better. It has to pay the same employee in the same currency. So the first question of the review isn't “why aren't they using it” but “what does the person do instead, and what do they get out of it.” ## Why does an immediate benefit beat a delayed one? Kaitlin Woolley and Ayelet Fishbach of the University of Chicago showed in a series of five studies that persistence in long-term goals is better predicted by immediate rewards than by delayed ones. People take up exercise or studying for the distant result, but they keep going because of what feels good now. The pattern held over a one-week horizon and over a three-month one. Parrott uses this to explain why a writing plugin stuck with her instead: even before it started working well, it gave her a paragraph to push off from, or a question that got the argument moving. An automation that saves ten minutes today survives more often than one that improves the quarterly report. That doesn't make reporting any less necessary; it's just a losing place to start: the payoff comes in three months, and the habit is forming now. ## Why does a written procedure strengthen motivation but not create it? The standard answer to “they're not using it” is to write a procedure. A meta-analysis by Paschal Sheeran, Olivia Listrom and Peter Gollwitzer in the European Review of Social Psychology pooled 642 independent tests of if-then plans. Such plans do work: effect sizes range from 0.27 to 0.66 depending on the type of outcome. The effect is larger when the plan has a strictly conditional format (“if X happens, then I do Y”), when the person is already motivated to reach the goal, and when the plan is said out loud. The key condition is “when the person is already motivated.” Planning strengthens a desire that already exists. It doesn't make something necessary if the person doesn't need it. A procedure will help a manager who wants to see the pipeline and forgets to update statuses: there's a goal, there's forgetfulness, there's a moment to anchor the action to. It won't help where updating a status gives the rep nothing but an extra step between two calls. ## What are the four reasons an automation dies? Going through her own abandoned systems, Parrott boiled the reasons down to four. In CRM practice they look like this: - It was forgotten. The automation exists, but it isn't in front of the person at the moment it's needed. - It didn't run. The trigger didn't fire, the integration broke, nobody saw the error. - Its output needs rework. Formally there is an output, but it takes so many fixes that doing it by hand is quicker. - The task doesn't recur. The problem comes up once a quarter, and a habit has nothing to build on. None of these four reasons is cured by training the team. The first two are engineering: visibility at the moment of need and failure monitoring. The third is output quality. The fourth means the automation shouldn't have been built at all, and the most honest decision here is to switch it off. ## Which three rules make the decision for you? From this analysis Parrott drew three working rules for herself. A new automation has to run manually three times before it goes on a schedule. Its output has to be usable after five minutes of review, no more. Four unused runs in a row are a reason to redesign it or switch it off. Each team sets the thresholds to its own rhythm; what matters is that they're stated before launch. Then “let's wait a little longer” stops being the default strategy, and the argument about an automation's fate stops being an argument about who's to blame. Auspex applies the same principle during implementation: what exactly should come out and how we'll know it did is written down before the system is configured, not after launch. ## How do you review your automations in one day? A review doesn't need an outside contractor and fits into one working day. Step one is an inventory. Write down everything that runs or is supposed to run: automation rules and triggers in the CRM, scenarios in Make or n8n, chatbots, automated mailings, AI assistants. For each item, note the owner and the date it last ran. The “owner” column regularly turns out to hold someone who left last year, while the scenario is still writing something every night. Step two is four questions for each item on the list: - What specific task was it supposed to handle, and whose task is it? - How many times did it run in the last month? - What does the person do instead, and what do they get out of it? - How long does it take to get the output into usable shape? Step three is a decision from four options: keep, redesign, postpone, switch off. “Postpone” needs to be a category of its own: some automations fail not because they're bad but because the process they were built for hasn't settled yet and the context is different every month. Dead automations aren't neutral. They keep writing data, corrupt the reporting and create the feeling that the process is under control. Switching one off is an action, not a refusal to act. ## What are the signs of an automation that will stick? Put the factors from the three studies together with the four reasons for abandonment and you get a checklist for judging an idea before it's built: - It sits inside work the person already does every day, not next to it. - It was requested or chosen by the person who will use it. - It gives back something useful the same day, not at the end of the month. - Its output is usable after a short check. - It's triggered by a stable event in the workday, not by an intention to remember. - You can see when it didn't run. If an idea fails on four of the six points, it's cheaper to redesign it on paper than to rescue it with training later. ## What does Auspex do during implementation to avoid building a graveyard? Auspex implements CRM and automates business processes for small and mid-sized businesses. The way we work matches what these studies show at three points. First: we start with a conversation about the process, not a demo of the system, and we don't automate a broken process, because the automation would inherit the breakage. Second: we take the boring, repetitive tasks first — invoices, reminders, an instant reply to an inbound lead. They pay back the same week and they come up often, which are exactly the two conditions behavioral science calls decisive. Third: after launch we stay in touch and look at what actually gets used — that's where you see which scenario has gone quiet. What we don't do: we don't judge whether an automation has stuck two weeks after launch, and we don't put non-use down to the team's discipline alone. ## Where do you start this week? - Write down every automation you have and the date each one last ran. - Mark the ones that have been silent for more than a month and ask the four questions from the protocol above about each. - Switch off whatever didn't pass the check, without putting it off “until next time.” The most expensive part of this review is permission to switch things off — especially things that took a long time and real money to set up a year ago. The analysis here is easier than the decision. If the inventory leaves you with a “redesign or switch off” list, bring it to a CRM and automation audit: we'll go through the process and give you a straight answer on every item. Read next - [Where to start automating in a small business: 5 boring processes](/en/blog/where-to-start-automation-smb/) - [Automations that pay for themselves in the first month](/en/blog/automations-that-pay-for-themselves/) - [The AI champion: who actually carries the rollout](/en/blog/internal-ai-champion/) - [How we roll out a CRM: eight steps](/en/blog/how-auspex-rolls-out-crm/) - [CRM and automation audit](/en/services/crm-audit/) ## Sources - [Every — Why Some AI Workflows Stick — And Others Don't (Katie Parrott)](https://every.to/working-overtime/why-some-ai-workflows-stick-and-others-don-t) - [Singh et al. (2024), Time to Form a Habit: A Systematic Review and Meta-Analysis — Healthcare 12(23)](https://doi.org/10.3390/healthcare12232488) - [Healthcare (Basel) 2024 — Time to Form a Habit: A Systematic Review and Meta-Analysis](https://pmc.ncbi.nlm.nih.gov/articles/PMC11641623/) - [European Review of Social Psychology — The when and how of planning: meta-analysis of 642 tests](https://doi.org/10.1080/10463283.2024.2334563) - [Personality and Social Psychology Bulletin — Immediate Rewards Predict Adherence to Long-Term Goals](https://pubmed.ncbi.nlm.nih.gov/27899467/) - [Scientific American — How Long Does It Really Take to Form a Habit?](https://www.scientificamerican.com/article/how-long-does-it-really-take-to-form-a-habit/) Andrew Maryasov Founder, Auspex Founder of Auspex. Implementing CRM and automating business processes since 2015 — 1,500+ projects for 1,200+ companies. [amaryasov.expert →](https://amaryasov.expert) ## Frequently asked How long before you can tell whether an automation has stuck? Two weeks is too early. A 2024 systematic review in the journal Healthcare recorded habit-formation medians of 59 and 66 days, with individual variation from 4 to 335 days, and that data is about everyday behaviors, not work processes. It's more practical to go by behavior than by the calendar: how many times the automation ran in a month, and in how many of those runs its output was actually used. Is it true that a habit forms in 21 days? No. The number comes from Maxwell Maltz's book Psycho-Cybernetics and described observations of plastic surgery patients. Research data gives very different figures: medians of about two months, means of 106 and 154 days, a range of 4 to 335 days. In one of the studies only 23% of participants reached the set automaticity threshold. Why doesn't the team use the CRM when everything is set up? Most often because the action in the system competes with an existing habit that gives the employee something tangible: speed, control, personal contact. Calling instead of logging it in the record saves a minute now, while the benefit of a filled-in pipeline arrives someday and goes to someone other than the person who filled it in. Until that asymmetry is removed, training has only a temporary effect. Will a written procedure help if an automation is being ignored? Partly, and on one condition. A meta-analysis of 642 tests of if-then plans found effect sizes from 0.27 to 0.66, and the effect is larger when the person is already motivated to reach the goal. A procedure strengthens a desire that already exists and fights forgetfulness. It doesn't create a need where there isn't one. Which automations stick best? The ones that sit inside daily work, pay back the same day, run often and don't need their output reworked. An additional factor is personal choice: in the studies, habits people chose for themselves formed more strongly than habits assigned from outside. That's why an automation a rep asked for has better odds than one handed down from above. Should you switch off an automation nobody uses? Yes, if it didn't pass the review. Scenarios that don't work aren't neutral: they keep writing data, distort the reporting and create a false sense of control over the process. Switching one off is a decision that frees the team's attention for the automations that really work. How can you tell a process isn't ready for automation yet? The main sign is that the task doesn't recur, or is done differently every time. A habit forms by repeating an action in a stable context, so a process that comes up once a quarter and looks different each time provides neither frequency nor context. In that case you describe and stabilize the process first and automate it afterwards. ## Business automation digest 2–3 emails a month — what actually works in CRM and automation. No spam. Unsubscribe in one click. [Free CRM audit →](/en/contact/) ↳ FREE AUDIT 30 min · free · ↓ ## Ready to Get Your Sales Under Control? Free process audit — 30 minutes, and you'll receive: 1,200+ companies since 2015 · 30-day median go-live 01 A map of where you're losing leads 02 Priorities: what to automate first 03 An implementation plan for 2-4 weeks We'll call within 24 hours. The plan is yours, even if we don't work together.